X blocks Grok from generating revealing images of real people after deepfake backlash

[midian182]

What just happened? X has announced that Grok's ability to edit images of real people into revealing clothing such as bikinis has been removed – even for paying subscribers. But there are reports of this restriction being simple to bypass.

"We have implemented technological measures to prevent the Grok account from allowing the editing of images of real people in revealing clothing such as bikinis. This restriction applies to all users, including paid subscribers," wrote the @safety account on X.

The post adds that image creation and the ability to edit images via the Grok account on the X platform are now only available to paid subscribers.

UPDATE FROM 𝕏 SAFETY TEAM

𝕏 has shared new safety updates regarding Grok image generation

New Grok image safeguards:
• Editing images of real people in revealing clothing (e.g. bikinis) is now blocked
• Applies to all users, including paid subscribers
• Image creation &… pic.twitter.com/JUOve9iGhe

– X Freeze (@XFreeze) January 15, 2026

The pressure on X, xAI, and Elon Musk has been growing more intense since the revelation that thousands of nonconsensual sexual deepfakes of women and sometimes children were being created on X via the Grok account. A report showed that 6,700 of these images were being generated on the platform every hour.

X initially said it was addressing the problem by restricting access to Grok's image editing feature, but this only stopped free users' requests when publicly replying to the Grok chatbot. Image editing remained available on the desktop site, while app users could access the same feature by long-pressing on an image. There were no changes to the standalone app, either. The latest update supposedly stops the Grok account from editing images of women in this way.

The @safety account adds that X will geoblock the ability of all users to generate images of real people in bikinis, underwear, and similar attire via the Grok account and in Grok on X in regions where it's illegal.

Can anyone actually break Grok image moderation? Reply below. https://t.co/4Dvj3RNyU5

– Elon Musk (@elonmusk) January 14, 2026

One place where that geoblocking applies is the UK. X said it was doing this to comply with a law that criminalizes the creation of nonconsensual deepfake images. Attempting to use Grok for this purpose in the country results in a message that reads, "Unfortunately I can't generate that kind of image."

The UK government has claimed "vindication" after the most recent announcement. Media watchdog Ofcom said it was a "welcome development," but added that its formal investigation remains ongoing.

However, The Verge reports that in the UK, undressing images of real women and editing them into sexualized poses is still possible using the X mobile apps or websites – using free X and Grok accounts – often by just using different phrasing. It can also generate images of men and inanimate objects in bikinis.

xAI and X are now facing an investigation into the deepfake images by California's attorney general. There have also been calls from three Democratic senators for Apple and Google to remove the apps from their stores as they violate the companies' respective policies.

Musk previously said that anyone using Grok to make illegal content would suffer the same consequences as if they uploaded it themselves. He also denied that the chatbot had been used to make images of children. "I [sic] not aware of any naked underage images generated by Grok. Literally zero," Musk wrote, hours before the latest change.

Image credit: Salvador Rios

Permalink to story:

X blocks Grok from generating revealing images of real people after deepfake backlash

 

[Xelions]

Has anyone actually compared the AI nude image to the actual? and look at the accuracy of such data... Could always use a specific AI model and input various clothed porn stars and compared them to their naked selves, in detail and for both genders. I hardly think a person at any age is gonna be represented accurately through AI generation.

What else comes to mind is that through Internet history there have been deep fakes, photoshopped nudes, whatever else you can think of - it's all been done. Only difference today is it's more automated. Governments cracking down on this is hilarious at best and almost controlling at worst. If someone wants to nudify a pic on their personal computer with their own AI model - who would know and to be honest who cares. It's when it starts getting attached to a name, a person, an account, and spreading online when it can affect someone, assuming it's even remotely accurate. That's only when it becomes a problem.

 

[unoficialoficial]

There is no way to regulate a neural network regarding whether it knows anatomy or does not. There is no middle ground where it knows anatomy but pretends that it does not know; it does not know shame, it just thinks. If they don't teach it anatomy, not only will it not be useful, but it will make dangerous mistakes. When someone tries to regulate the neural networks (AI), they are basically regulating the training set. So, if you don't want nudity, you must be ready to say AI should not know about anatomy, and there should not be anatomy-related content in the training dataset. Because if it knows, when asked, it will express its thoughts. In India, maybe they want to apply a layer of ethics around cows, in Spain a completely opposite layer of ethics around cows. That thing is problematic because AI is universal, not local, so it should be neutral. It is impossible to obey all laws worldwide without becoming rigid and useless.